08.23.2004 [asutton@cs.kent.edu]
	o Rewrote OMF::Set to use std::map instead of std::set.
	  This actually makes the OMF::Set class searchable by
	  without having to rely on pointer values. It uses the
	  object's hash code instead.
	o Rewrote all singletons to be statically allocated and
	  destroyed. This was originally intended to improve memory
	  deallocation on exit, but ended up causing bugs in the
	  MOF implementation. This change may get reverted at a
	  later date.
	o Singleton access functions now return references instead
	  of pointers. This prevents the notion that users can delete
	  singletons.
	o Completely rewrote the python bindings to use real wrapper
	  structs instead of what I'd done before.
	o Python Model.Package is now derived from OMF.Package and
	  usable as such (it's a nice interface fix).
	o Seem to have fixed the stack squashing bug I had in the
	  previous release (as create.py now executes completely).
	o Seem to have fixed the "pure virtual" exceptions I was
	  getting. Note to self: always incref() boost.python
	  objects before they go out of scope.
	o Started writing regression tests to exhaustively test
	  the OMF and MOF interfaces. Probably about 30% complete
	  by now.

08.24.2004 [asutton@cs.kent.edu]
	o Fixed some XMI parsing bugs where I wasn't using the
	  association extents to set associations, I was just
	  using the property. Now, it actually works right.
	o Fixed Model::StructuralFeature destructor to actually
	  get rid of the contained multiplicity.
	o Fixed another Python bug that caused crashes because
	  of poorly managed memory in the bindings.

08.26.2004 [asutton@cs.kent.edu]
	o Added functionality to switch the parsing mechanims
	  for XMI reading. Renamed old XmiReader to XmiReader_2_0.
	o Added a class for parsing XMI 1.2. I basically did this
	  to get a quick and dirty UML metamodel. This class is
	  still pretty experimental and doesn't - to my knowledge
	  handle all cases of XMI production.
	o Added ordered property retrieval to OMF::ModelObject to
	  support <XMI.field/> elements in XMI 1.2.
08.27.2004
	o Made OMF::XmiReader_1_2 a lot more stable. It now works
	  regardless of the location of the xmi.id and appears to
	  resolve all the forward references.
	o Fixed a property definition in Model::Constraint. Needed
	  to list "constrainedElement" in the first position.
08.28.2004
	o Fixed a bug in OMF::ModelService that failed to return
	  models after they'd already been loaded
	o Fixed a bindings bug where Py_None wasn't being incref'd
	  before returning. This would cause Python to release the
	  Py_None instance, aborting a process.
09.07.2004
	o OMF2 gets UML! Finished the code generator to generate a
	  pretty basic UML implementation. Pending massive faults
	  in the generation, this should stay pretty stable.
	o Moved some Python support into the OMF core to support
	  actual metamodel class types - no more OMF.Elements.
09.08.2004
	o Generated UML extensions for the Python bindings. Now, the
	  OMF2 is theoretically back in line with OMF1.